Running CultureGroup simulation

File overview

The following files are found in the methods/ folder

• sbatch.sh initates a job array on slurm
• run.jl initates julia simulations after loading methods from methods.jl and the job array specific parameters from main.jl
• summariseFiles.R opens all *.Rdata files from the output directory, summarises values, returns a summary file into the working directory, zips the output folder, deletes the output folder and finally deletes all slurm*.out files

Requirements

Running the simulation code requires:

• julia (dependencies: Distributions, StatsBase, RCall)
• r (dependencies: igraph, dplyr, reshape2)

Simulations reported in our article

In our article, Cultural Selection Shapes Network Structure (https://www.biorxiv.org/content/early/2018/11/08/464883), we report results for the following iterations of the model:

1. Social learning dynamics for fixed simple graphs
2. Social learning dynamics for dynamic complex networks with fixed linking parameters
3. Social learning dynamics for dynamic complex networks with evolving linking parameters
4. Social learning dynamics for complex networks with evolving but coupled linking parameters
5. Social learning dynamics for dynamic complex networks with switching selection regimes

Furthermore, in the ESM we report results for simulations with

6. Low mutation rate
7. Connection costs
8. Varying population size and trait number
9. Varying innovation and social learning success rate

To run the individual simulations follow the steps outlined below to adjust the simulation.

Adjust simulation

1. Social learning dynamics for fixed simple graphs

To use a simple graph such as a ring (as in the main text), find # Setup for evolution in the main.jl file and make sure evolveNetwork is set to true. This will which will create a ring graph of size nod with neighbourhood neibhood, and keep the network shape fixed. Note: this is currently only implemented for neutral selection, as parents for a newborn have to be adjacent to the newborn and are not selected based on their fitness.

Can be combined with: 8, and 9

2. Social learning dynamics for dynamic complex networks with fixed linking parameters

To let linking parameters evolve, find # Setup for evolution in the main.jl file and make sure evolvePN and evolvePR are set to false. This will keep the linking parameters fixed throughout a simulation, while the network is still dynamically rewired.

Can be combined with: 5, 7, 8, and 9

3. Social learning dynamics for dynamic complex networks with evolving linking parameters

To let linking parameters evolve, find # Setup for evolution in the main.jl file and make sure evolvePN and evolvePR are set to true. This will let the parameters evolve (and mutate) throughout a simulation.

Can be combined with: 6, 7, 8, and 9

4. Social learning dynamics for complex networks with evolving but coupled linking parameters

Change grid parameter in line 18 (pnprcoupled) from false(not coupled), to true (couples pr to pn given an average degree k and population size N). In the main text we used average degrees 2, 6, 10, for a population of 100 individuals.

Can be combined with: 6, 7 (note, in this case mutation only affects PN), 8, and 9

5. Social learning dynamics for dynamic complex networks with switching selection regimes

To enable switching payoff regimes find ## 3EXPLOIT in methodsJL and uncomment the section starting with # Switching payoff method twice throughout the simulation. This will change the payoff regime twice throughout a simulation run (after 1/3 and 2/3 of the rounds).

Can be combined with: 6, 7, 8, and 9

6. Low mutation rate

Change grid parameter in line 19 (mutation rate, mutRate). We used 1 for the main text and 0.01 for the appropriate simulations in the ESM.

7. Connection costs

Change grid parameter in line 17 (connection cost, cc). We used 0 for the no cost and 0.01 for the cost condition.

8. Varying population size and trait number

Change grid parameter in line 1 (population size) and line 4 (number of seed traits). In the main text we used 100 for each parameter.

9. Varying innovation and social learning success rate

Change grid parameters in line 2 (social learning success rate, socialLearningSuc) and line 14 (individual learning success rate, indSuccessRate). In the main text we used 0.75 and 0.01 respectively.